Carnegie Mellon University v. LSI Corporation — Court Excludes Patent Damages Expert Over Time-Bar and Apportionment Failures

Background

Carnegie Mellon University (CMU) holds two related patents covering improved methods for accurately reading bits stored on hard disks. CMU sued LSI Corporation and Avago Technologies, alleging that LSI’s read channel system-on-a-chip (SoC) products—sold to customers Western Digital, Toshiba, Seagate, and Hitachi Global Storage Technologies (HGST)—infringed its patented methods by practicing them millions to billions of times per second during testing, validation, and operation.

Because CMU asserts method claims rather than system claims, direct infringement arises from the use of the claimed methods, not from selling the products. Nevertheless, CMU based its damages on LSI’s product sales, arguing that without the patented methods, LSI would never have reached volume production. The Federal Circuit accepted a similar approach in CMU’s earlier $1.5 billion verdict against Marvell Technology involving the same patents.

A critical wrinkle: Seagate and HGST held licenses to CMU’s patents with have-made rights, meaning LSI was authorized to manufacture the patented technology for those customers. The court had previously ruled on summary judgment that LSI’s sales to Seagate and HGST fell within these have-made rights. LSI moved to exclude CMU’s damages expert, Catharine Lawton, under Federal Rule of Evidence 702.

The Court’s Holding

Judge Donato partially excluded Lawton’s damages opinions on two independent grounds.

Time-barred infringement. For Seagate and HGST (both licensees), CMU could not pursue indirect infringement theories because a licensee’s use of patented technology is authorized and thus not direct infringement—a prerequisite for indirect liability. CMU’s remaining theory was that LSI directly infringed during a “customer-independent” platform development stage before have-made rights applied. But this development occurred years before LSI’s product sales. Because the statute of limitations under 35 U.S.C. § 286 bars recovery for infringement committed more than six years before the lawsuit, and the infringing acts (development) preceded the sales (the royalty base) by a substantial period, Lawton’s failure to separate time-barred development-stage infringement from timely sales rendered her HGST and Seagate opinions methodologically flawed.

The court emphasized a key distinction between method claims and system claims: for system claims, the sale itself is the infringing act, so the infringement and the revenue it generates are roughly contemporaneous. For method claims, the infringing use may occur long before any sale, creating a time lag that a damages expert must account for under § 286.

Failure to apportion. Lawton attributed the entire 0.6 dB signal-to-noise ratio (SNR) improvement between two LSI product generations to the patented technology. But CMU’s own technical expert conceded that “some portion” of the improvement came from unpatented features like parity optimization. The Supreme Court’s longstanding apportionment requirement, from Garretson v. Clark (1884) through the Federal Circuit’s VirnetX and Finjan decisions, requires a patentee to separate damages attributable to patented versus unpatented features. Lawton’s failure to do so warranted exclusion of her Redtail-generation SNR opinion.

The court allowed Lawton’s opinions regarding sales to Western Digital and Toshiba to stand. Those customers were not licensees, so all three infringement theories (direct, induced, and contributory) applied, and the time-lag problem did not arise because contributory infringement occurs at the point of sale.

Key Takeaways

• Method claims create a unique damages timing problem. Unlike system claims where the sale is the infringing act, method claim infringement (use during development) and the resulting revenue (later sales) may be separated by years—and the statute of limitations applies to the infringement date, not the sale date.
• Have-made rights can dramatically limit a damages theory. When a customer holds a patent license with have-made rights, the patent holder cannot pursue indirect infringement theories against the manufacturer, and must rely solely on direct infringement during development—which may fall outside the limitations period.
• Your own expert can sink your damages case. CMU’s technical expert’s concession that unpatented features contributed “some portion” of a performance improvement created an apportionment obligation that the damages expert failed to satisfy.
• Prior victories in related cases do not guarantee the same approach works again. CMU’s $1.5 billion Marvell verdict used a similar methodology, but materially different facts (Seagate’s license status) made the precedent inapplicable here.

Why It Matters

This ruling highlights a frequently overlooked challenge in patent damages for method claims: the temporal disconnect between when infringement occurs and when revenue is generated. In an era of increasingly complex technology supply chains—where chip development, manufacturing, and sales span years—patent holders asserting method claims must carefully align their damages theories with the statute of limitations, especially when some customers have independent patent rights. The decision also reinforces that the apportionment requirement has real teeth: even a modest concession by a technical expert that unpatented improvements contribute “some portion” of a performance gain is enough to exclude a damages opinion that treats the entire improvement as attributable to the patented invention.